**Use adjusted close for a simple backtest based on percentage returns. Use actual historical prices plus explicit corporate-action accounting for a backtest that simulates trades and portfolio cash.**

The two prices answer different questions:

| Data | What it represents | Typical use |
|---|---|---|
| Unadjusted close | The closing market price that day | Trade prices, share quantities, dollar-based rules |
| Adjusted close | Historical prices rescaled to account for splits and, usually, dividends | Comparing performance across time, calculating returns |

Check your provider’s definition. For example, Alpha Vantage supplies raw prices, adjusted close, and separate dividend and split events. [Alpha Vantage documentation](https://www.alphavantage.co/documentation/)

Two examples explain why adjustments matter:

- **Split:** You own one share worth $100. A 2-for-1 split leaves you with two shares worth $50 each. Your holding is still worth $100. Calculating returns from unadjusted closes alone would incorrectly show a 50% loss.
- **Dividend:** Suppose a $100 stock pays a $2 cash dividend and its price falls to $98, with everything else unchanged. You hold $98 of stock plus $2 cash. A price-only calculation shows a 2% loss; your total wealth is unchanged.

Adjusted prices make these observations comparable. **Their historical dollar levels are accounting values, so they should not be treated as prices you could have traded at.**

For your backtest:

- **Simple daily return or momentum research:** Adjusted close is generally a useful starting point. Check the provider’s dividend methodology and reinvestment assumptions.
- **Trade and portfolio simulation:** Use raw prices for fills and position sizing. Adjust share quantities for splits and credit dividends to eligible holdings, with cash available on the payment date. QuantConnect documents this approach in its raw-data mode. [Corporate-action handling](https://www.quantconnect.com/docs/v2/writing-algorithms/securities/asset-classes/us-equity/corporate-actions)
- **Signals:** Split-adjusted prices prevent artificial jumps. Whether to adjust for dividends depends on whether your signal measures price movement or total return.

Two pitfalls matter: **don’t count dividends twice** by both using dividend-adjusted returns and adding dividend cash; and use historical market prices for rules such as “buy stocks below $10.” Later corporate actions can change back-adjusted price levels and introduce look-ahead bias into those rules. [QuantConnect research guide](https://www.quantconnect.com/docs/v2/writing-algorithms/key-concepts/research-guide#08-Look-ahead-Bias)